A crystal decanter of amber whiskey, two lowball tumblers, a small mirror tray, a gold chain. 

A LEGO minifigure of Doug Ford, puffy, flushed red face, sandy blonde combed-over hair. 

Miami Vice meets Etobicoke.

That’s how a burgeoning social media influencer created a scene from The Gravy Plane, an AI-generated music video lampooning Premier Doug Ford’s private jet fiasco.

To human eyes, the 400 words of prompts might be an incoherent block of text.

To a generative artificial intelligence program, however, they’re taken as clear instructions for how to create the next frontier in political warfare.

The Gravy Plane parody video, with its catchy country-music soundtrack and a depiction of Doug Ford in LEGO form, has racked up hundreds of thousands of views across YouTube, TikTok, Instagram, X and Facebook.

It combines criticism of the government’s dalliance into jet ownership with hidden Easter eggs, such as the premier’s phone buzzing with a court order to release his cellphone records, or a hospital bed in a hallway — the kind of detail only a dedicated consumer of Ontario politics would know.

So it may come as a surprise that its creator is a middle-aged man based in Switzerland.

Originally from Ottawa, Alex Hout relocated to Switzerland 25 years ago after quitting the Canadian restaurant business and now considers himself an AI storyteller.

“I happened to stumble on this story of the purchase of a $29 million plane, and then in a few days they announced they were selling it back, and I was like, ‘This story will write itself,’” Huot said, “and the person to write this story is me.”

While Huot estimates the Miami Vice meets Etobicoke scene inside the jet took a dozen tries to get right, the whole project took him a week.

“I write the text, I write the lyrics, and then from there I use tools to generate the music, images and videos,” Huot said. “It’s a bit of a dance between my idea and what the AI generates.”

In order to produce the videos, Hout estimates he spends a couple of hundred dollars on subscriptions to programs like Suno, Nano Banana and Higgsfield — all of it paid out of pocket.

Huot is adamant that he was not contacted by any political party to create both The Gravy Plane and a second catchy Ford critique called FOI, but the videos have garnered him attention from Queen’s Park.

Ontario NDP leader Marit Stiles, Interim Ontario Liberal leader John Fraser and Minister of Long-Term Care Natalia Kusendova-Bashta are all now among Huot’s followers on Instagram.

Huot also says he was approached by a school board trustee who asked if they could use the song in a meeting.

“Honestly, that’s quite humbling for me,” Huot said.

Ebrahim Bagheri, an expert in responsible AI use and professor at the University of Toronto, suggested it might level the political playing field.

“Typically, you would think about a team building such a video,” he said. “Now AI has enabled just one person [to do all of that].

“I think in the next year or so these tools will become democratized to the point where your average person has probably worked with or generated a bunch of these videos, at which point you’ll reach competitive parity.”

Bagheri sees the addition of generative AI to the campaign battlefield as a double-edged sword.

“The big concern is the authenticity of the content that’s being generated. The ones with animations and the Lego figures are obviously creative work, and people will realize that immediately.”

“However, on the other side, you have the deepfakes, where people generate videos of public figures.”

In early May, the City of Toronto shut down the Gardiner Expressway and Don Valley Parkway on a weekend that featured soccer superstar Lionel Messi coming to town, three Toronto Blue Jays games and Mother’s Day.

An organization called IntegrityTO posted a deepfake-style video to X, showing Chow uttering words she never said about closing the highways.

While portions of the video are satirical in nature, and the voice used for the mayor is clearly inaccurate, showing the mayor at a podium with her lips moving to match the fake speech was enough to seemingly fool some in the X comments.

The group insists it wasn’t trying to trick anyone.

“I think anybody with any degree of common sense would know that it’s AI, because it sounded nothing like her voice,” said Daniel Tate, IntegrityTO’s executive director, who likened the video to a newspaper caricature.

“Political cartoons are part of a free, democratic society, so what we’re seeing here is a very realistic drawing, a robot’s drawing in this case.”

Others — both self-identifying conservatives and Chow supporters — reacted poorly to the video.

“The deepfakes are becoming a substantial threat to the integrity of our society,” Bagheri said.

“It’s not just the political system. There are so many smaller pieces, societal pieces, that are being impacted by generative AI technology, the political system is only one of them.”

Stephen Taylor agrees. His company created Flashbulb, a media monitoring AI tool that amasses content from television, radio, print and parliament committees and distills it into a digestible format.

“It’s certainly a concern, and it’s upon all of us to become educated with what’s out there and how these techniques are being used,” Taylor said.

“It’s not just people making videos about the premier and his plane, it can be foreign interference and people trying to manipulate how we see domestic political issues from the point of view of an adversarial power, rather than just typical democratic participation among citizens within our own country.”

Taylor also sees AI in campaigning as something that’s here to stay, akin to other revolutionary digital tools like Photoshop or search engines.

“I think the typical voter and citizen will get more fluent with AI,” Taylor said. “They’ll develop an antennae.”

At Queen’s Park, the Ontario Liberals tabled a bill to protect against malicious artificially generated content, arguing the next election could be heavily influenced by AI, and that guardrails would need to be put in place.

“Basically, you can do just about anything, you can have anybody saying anything or doing anything, and that breaks down trust,” interim party leader John Fraser said. “All of a sudden, people don’t know if what they’re seeing is real.”

Despite emphasizing the bipartisan nature of the law, the Progressive Conservatives shot it down.

So, in the absence of any actual rules or legislation, what are the best practices of AI politics?

“One of the key things is disclosure,” Bagheri said and added that another is quality control.

“The onus remains with the [user] themselves to check the validity of the content that’s being created by AI. If there’s trouble, you cannot blame the AI model.”